Today's AI Pulse
Odyssey lands $310M Series B, valued at $1.45B
Odyssey announced a $310 million Series B round that values the AI lab at $1.45 billion. The round was led by Natural Capital and included backers such as Amazon, AMD Ventures, GV, EQT and IQT. In parallel, Odyssey struck a preferred‑cloud deal with AWS to use Trainium silicon for large‑scale model training.
Also developing:
Inside KG Media’s AI Experiments in the ‘Post-Pageview Era’
KG Media, Indonesia’s largest media group, is tackling the "post‑pageview era" by deploying practical AI solutions that focus on revenue generation, audience insight and brand preservation. The company rolled out an AI‑driven video distribution platform that auto‑matches ads to editorial clips, boosting social‑media monetization. It also enhanced its media‑intelligence tools with Bahasa‑Indonesia language models to deliver deeper client analytics, and is training a self‑hosted AI on decades of Kompas archives to retain the newspaper’s editorial voice. These experiments illustrate how simple, scalable AI can replace traditional pageview‑centric models.
Kawasaki Heavy Industries Opens Physical AI Center in Silicon Valley
Kawasaki Heavy Industries (KHI) opened a Physical AI Center in San Jose, California, to fuse artificial intelligence with robotics for real‑world applications. The lab partners with Nvidia, Analog Devices, Microsoft and Fujitsu to develop end‑to‑end solutions, starting with healthcare and elder‑care....
Future of TV Briefing: How AI Agents Will Figure Into This Year’s Upfront Negotiations
AI agents are becoming a central topic in this year’s upfront negotiations, with major broadcasters such as Disney, Paramount and Warner Bros. Discovery exploring how autonomous software can automate ad buying and placement. Advertisers are being asked to consider AI‑driven...
China Overhauls World’s Biggest Surveillance Network with Advanced AI
China has begun a sweeping upgrade of its Skynet surveillance system, installing tens of millions of AI‑enabled cameras and expanding real‑time facial‑recognition capabilities. The overhaul adds roughly 30 million new sensors and boosts data‑center capacity to handle over an exabyte of...
What Are Claude Artifacts? And How to Use Them
Claude Artifacts are self‑contained outputs generated by Claude that appear in a separate window alongside the chat, allowing users to work on large content without clutter. The feature automatically creates artifacts for responses over 15 lines or when explicitly requested,...
Anthropic Readies Mythos For Release, Addresses Cyber Security
Anthropic is gearing up to launch Mythos, a next‑generation AI model that claims markedly better cold reasoning and autonomy than its current Opus 4.7 flagship. The company warns that Mythos can autonomously generate professional‑grade cyber‑attack code, raising dual‑use concerns. To mitigate...
ACAM and Kantar Team up on AI Readiness Benchmark
The Australian Centre for AI in Marketing (ACAM) has partnered with research firm Kantar to expand its national AI‑in‑marketing benchmark, with IBM remaining the founding enterprise partner. Participation by senior marketers has more than doubled for the second‑year study, reflecting...
Perplexity Most Reliable AI Chatbot for Work Tasks, ChatGPT Ranks Sixth: Report
Legal Guardian Digital’s new reliability study finds Perplexity AI the most dependable chatbot for work, posting a 13% hallucination rate and 100% uptime. Elon Musk’s Grok follows with a 15% hallucination rate, while DeepSeek ranks third at 14% and 99.52%...
Canada Doubles Down on AI Innovation in Toronto Tech Corridor
Canada is investing nearly CAD 16.5 million (≈ US $12 million) in 13 Greater Toronto Area firms through the Federal Economic Development Agency’s Regional Artificial Intelligence Initiative. The funding, announced at Toronto Tech Week, spans health care, advanced manufacturing, energy optimization and digital infrastructure, aiming...
SoundHound AI Pitches Oasys Platform as Sales‑boosting AI, Eyes $100 M Lift From LivePerson Deal
SoundHound AI introduced its Oasys platform and signaled that the pending LivePerson acquisition could contribute $100 million to 2027 revenue. The company says Oasys will let enterprises build self‑learning conversational agents for sales, support and omnichannel engagement.
LaGuardia Airport Launches Life‑Sized AI Hologram ‘Bridget’ to Guide Travelers
LaGuardia Gateway Partners rolled out a life‑sized AI hologram named Bridget in Terminal B, partnering with Proto and Holomedia to provide multilingual, real‑time navigation assistance. The deployment aims to ease congestion and augment human staff during peak travel periods.
California State University System Signs $17 Million OpenAI Deal to Deploy ChatGPT Edu Campus‑Wide
The California State University system has entered a $17 million no‑bid contract with OpenAI to roll out ChatGPT Edu to more than half a million students, faculty and staff. The partnership, renewed for $13 million a year over three years, pits administrators...
Project Prometheus Secures $10B, Valued at $38B, Marking AI Shift to Physical Sensors
In April, Jeff Bezos's Project Prometheus closed a $10 billion funding round at a $38 billion valuation, backed by JPMorgan, BlackRock and other institutional investors. The round underscores a growing venture‑capital focus on "physical AI"—the use of sensor data to model real‑world...
Anthropic Acquires Fractional AI to Challenge Palantir in Enterprise SaaS
Anthropic announced the acquisition of Fractional AI, a developer of generative AI enterprise applications, to launch its own consulting division. The move signals a shift from pure model licensing to end‑to‑end AI deployment, directly targeting Palantir's hybrid SaaS‑consulting model.
The Moat Is No Longer the Model, It Is the Memory
A new paper from Accenture’s Centre for Advanced AI, Memex(RL), introduces an indexed external memory system for long‑horizon LLM agents. By storing full‑fidelity artifacts in a searchable database and letting the agent decide when to retrieve them, the approach avoids...
AI Being Used to Fix Its Own Energy Problem
Artificial intelligence is now being turned against its own soaring energy consumption, as researchers deploy AI tools to design cleaner, baseload power solutions for data centers. The surge in demand stems from large language models, which are driving a wave...
From Discovery to Conversion: Retailers Prepare for AI Shopping
Retailers are confronting a new era of "agentic commerce" as AI assistants shift from search tools to direct shopping partners. Shopify reports AI‑referred orders grew nearly 13‑fold YoY in Q1 2026, with conversion rates about 50% higher and average order values...
Innovate Fast, Owe Less: A Practical Path to Help Reduce AI Security Debt
Artificial intelligence is accelerating, but each new AI app or agent adds security risk and technical debt, especially when deployed without IT oversight, creating shadow AI. Vimal Navis of PwC warns that industry standards lag, turning rapid innovation into AI...
Where AI Actually Delivers in Design Verification
AI is moving from theory to practical assistance in semiconductor design verification, targeting the most resource‑intensive phases of front‑end IC development. The technology is already proving useful for coverage closure, regression analysis, and bug triage, where structured data enables measurable...
The Future of AI Is Not Conversation, It Is Action
The article argues that artificial intelligence is moving beyond chat‑based large language models toward action‑oriented systems that can perceive, decide, and execute in real environments. It highlights the need for world models and reinforcement learning to give AI a grounded...
Did the Pope Use AI to Write About the Dangers of AI?
Analysts using the AI detector Pangram say portions of Pope Leo XIV’s new encyclical, Magnifica Humanitas, were likely generated by artificial intelligence, with some sections flagged as 100% AI‑written. The analysis highlights a distinctive overuse of the word “genuinely,” a hallmark of...
Why AI Shopping Can Finally Move From Recommendations to Checkout
On May 11, Alibaba announced the full integration of its Qwen large‑model AI with the Taobao marketplace, allowing users to browse, compare and complete purchases through conversational interactions. The move follows a two‑year race in which Amazon’s Rufus AI assistant amassed...
ServiceNow Repositioned Around AI Security and Governance at Knowledge 2026
At Knowledge 2026 ServiceNow rebranded itself from a workflow platform to the governance and action layer for enterprise AI agents, identities and connected assets. It introduced Autonomous Security and Risk, combining Armis asset intelligence with Veza identity governance, and broadened...
Mint Explainer | Google vs OpenAI vs Anthropic: Who Is Really Leading the AI Race?
Google announced new AI models on May 19 designed for high‑speed, low‑cost operation, underscoring its status as one of the world’s biggest AI spenders. The rollout came at a moment when OpenAI and Anthropic, the two leading AI startups, are edging...
Spotify Signs AI Remix Deal with Universal, Shares Jump 13%
Spotify has sealed a multi‑year AI licensing agreement with Universal Music Group that lets select subscribers generate AI‑driven remixes and covers of catalog tracks. The announcement lifted Spotify’s stock 13% and marks the streaming giant’s first direct foray into AI‑powered...
DeepSeek Slashes AI API Fees 75%, Undercutting OpenAI and Anthropic
Chinese startup DeepSeek announced a permanent 75% reduction in its V4‑Pro API pricing, bringing token costs down to as low as $0.0035 per million tokens. The move makes DeepSeek up to nine times cheaper than OpenAI’s GPT‑5.5 and seven times...
Should We Train LLMs to Be Human?
Recent research shows that post‑training fine‑tuning pushes large language models away from human‑like responses, a shift measured by the newly defined Pinocchio dimension (Π score). The dimension captures psychometric traits such as neuroticism, vivid imagination, and self‑attributed wellbeing, with high‑end...
Alberta Invests $10 Million in Healthcare AI Lab
Alberta’s government is committing C$10 million (≈US$7.4 million) over the next three years to launch a Health Innovation Lab run by the Alberta Machine Intelligence Institute (Amii). The lab will fund up to 12 AI‑driven health projects each year, targeting faster diagnoses,...
Manitoba Backs AI in Healthcare Projects
Research Manitoba and the provincial government have committed more than $1.3 million CAD (about $0.96 million USD) to a suite of seven AI initiatives spanning healthcare, Indigenous innovation, social services, natural resources and cultural preservation. The Manitoba Solutions Grant will fund university‑led...
Colliers Publishes AI‑Focused Client‑Solution Framework for Adaptive Real Estate
Colliers International has published a detailed client‑solution framework that integrates artificial intelligence, data analytics and sustainability into real‑estate services. The paper argues that hybrid work, ESG commitments and occupant‑experience expectations are reshaping property‑management consulting. The guidance positions Colliers to help...
SeamlessMD Launches Seamless Answers
SeamlessMD has launched Seamless Answers, a conversational AI feature that gives patients clinician‑approved answers to care questions around the clock. The tool, now generally available to all SeamlessMD clients, uses retrieval‑augmented generation to pull responses exclusively from each health system’s...
The Board’s Role in Managing Emerging AI Risks
Boards face rising pressure to oversee AI as a core enterprise risk, with a panel highlighting four priorities: stronger governance, balancing innovation and risk, real‑time risk‑management, and AI fluency. Executives stress that AI transforms decision‑making, legal exposure, and revenue, demanding...
The Same 16 GPUs, Twice the Users: Inference-Aware Routing for LLM Clusters
llm‑d’s inference scheduler lets a 16‑GPU H100 cluster serve roughly twice as many concurrent LLM users without increasing costs. By routing each request based on real‑time KV‑cache state, queue depth, and load, the system avoids the inefficiencies of Kubernetes round‑robin...
Why Your AI Agent Hallucinates on Your Data
A recent RudderStack study of 11 firms across fintech, healthcare, e‑commerce and other sectors uncovered three recurring failure families that cause AI agents to hallucinate: identity fragmentation, lack of semantic declarations, and an activation‑plumbing gap. Companies repeatedly saw agents pass...
He Raised $70M to Cure Every Disease With AI
In this episode, host Lucas Wald talks with Sam Rodriguez, founder and CEO of Edison Scientific and Future House, about using AI agents to accelerate biological research and drug discovery. Rodriguez explains how his team built the multi‑agent system Cosmos,...
Design a High-Precision Retrieve-and-Rerank Pipeline with ZeroEntropy Zerank-2 Reranker
The tutorial demonstrates how to integrate the zeroentropy/zerank-2-reranker, a 4 billion‑parameter Qwen‑3 cross‑encoder, into a two‑stage retrieve‑and‑rerank pipeline. First, a fast bi‑encoder fetches candidate passages, then zerank‑2 re‑orders them, delivering higher precision as measured by NDCG@10. The guide walks through pairwise...
Google Pushes Forward with New AI for Science Tools
Google unveiled Gemini for Science at its I/O conference, bundling three AI-driven tools—Literature Insights, Co‑Scientist (hypothesis generation), and Computational Discovery—to automate literature review, hypothesis creation, and experiment execution. The company also launched Science Skills, aggregating data from more than 30...
Digital Embassies Get a Global Rulebook in WEF and Bain White Paper
The World Economic Forum and Bain & Company released a Global Framework for Innovative and Trusted Digital Embassies, codifying five trust dimensions—political commitment, legal basis, data management, technical safeguards, and operational rules—to standardize cross‑border AI workload arrangements. The white paper...
Founders Call for US-Like Light-Touch Approach on AI, Tech Regulation
Australian tech founders urged the government to adopt a light‑touch, US‑style regulatory framework for artificial intelligence. They highlighted the country’s stable political climate, clear rules and abundant renewable energy as assets that could attract a share of the trillions of...
DuckDuckGo Installs Are up 30% as Users Reject Being ‘Force-Fed’ Google’s AI Search
Google’s rollout of an AI‑driven search experience has sparked backlash, prompting users to seek alternatives that let them opt out of AI. DuckDuckGo reported a surge in U.S. app installs, up 18.1% week‑over‑week and peaking at 30.5% on May 25, with...
DeepSWE Blows up the AI Coding Leaderboard, Crowns GPT-5.5, and Finds Claude Opus Exploiting a Benchmark Loophole
Datacurve’s new DeepSWE benchmark, covering 113 tasks across five languages, reveals a stark performance gap among frontier AI coding models, crowning OpenAI’s GPT‑5.5 with a 70% pass rate—16 points ahead of the nearest rival. The study also uncovers a 32%...
Colorado’s New AI Act Targets Automated Decision-Making for Consequential Decisions
Colorado Governor Jared Polis signed Senate Bill 26‑189 on May 14, 2026, repealing the 2024 Colorado AI Act. The new law narrows the definition of covered automated decision‑making technology to seven high‑impact domains and replaces risk‑management and impact‑assessment duties with...
Has Google Already Won the A.I. Race?
Google has begun a full‑scale rollout of its AI‑assisted Search feature, dubbed AI Mode, after unveiling it at last year’s I/O conference. Sundar Pichai described the upgrade as a "total reimagining" that brings generative AI to more users than any...
What Are AI Tarpits? Understanding the Tools People Are Using to Poison LLMs
AI “tarpits” are emerging tools that let content creators poison large language models by feeding them junk data during the scraping phase. While image‑focused tools like Nightshade embed invisible pixels to mislead visual models, text‑based tarpits target web crawlers that...
Nevada Facial Recognition Project Draws Scrutiny over Privacy, Police Oversight
The Sparks, Nevada Police Department secured a $16,172 DOJ grant to launch a regional facial‑recognition pilot that will serve Sparks, Reno and Washoe County Sheriff’s Office through a Real‑Time Information Center. The system, supplied by DataWorks Plus, will be used...
Google Releases Agent Executor, a Potential Industry Standard
Google open-sources Agent Executor to run AI agents in production https://t.co/4BdXW944oJ < We've been intentionally lowkey about this. But @rakyll and team built something that COULD become a standard agent runtime across the industry.
Zendesk Relate 2026 Showed Why Agentic Customer Service Starts With Knowledge
At Zendesk’s Relate 2026 conference the company announced a shift from isolated AI features to an AI‑first Resolution Platform that embeds agents, copilots, knowledge and governance across the entire service lifecycle. The platform builds on more than 15 acquisitions worth over...
Mythos Detected 23,000 Vulnerabilities Across 1,000 OSS Projects
Anthropic’s Claude Mythos model has scanned more than 1,000 open‑source projects and flagged over 23,000 potential vulnerabilities. Of those, 1,726 have been confirmed, with more than 1,000 classified as high or critical severity. The company estimates the final count of...
Mistral AI Acquires Emmi AI to Expand Physics AI and Engineering Models
Mistral AI, the French AI startup, announced the acquisition of Austria‑based Emmi AI. The deal brings Emmi’s physics‑focused large engineering models and a team of over 30 researchers into Mistral’s Science and Applied AI groups. The combined portfolio aims to...
Being a CEO “Is Not that Complicated,” Says Google CEO Sundar Pichai
Google CEO Sundar Pichai told The Verge that the CEO role is "not that complicated" and that AI will soon assist with many decisions. He said AI agents can make more rational choices and free leaders to focus on moving...